TL;DR

  • We introduce a framework for harnessing Generative AI to track transnational repression at scale.
  • Early stage research proposal

The mobilizing capacity of diaspora dissent

Across highly visible mediums, prominent in anti-war/repression/apartheid movements

The role of diaspora dissidents in international politics

🏡

🌏

“Horizontal” networks at home
(Brinkerhoff 2009; Alonso and Oiarzabal 2010; Bernal 2020)

“Vertical” networks abroad
(Keck and Sikkink 1999; Michaelsen 2018; Esberg and Siegel 2020)

Strengthened ties with ⬆️ global migration, digitalization

🏡

🌍

“Horizontal” networks at home
(Brinkerhoff 2009; Alonso and Oiarzabal 2010; Bernal 2020)

“Vertical” networks abroad
(Keck and Sikkink 1999; Michaelsen 2018; Esberg and Siegel 2020)

Put yourself in the shoes of a dictator…

…are you going to allow this?

🏡

🌎

Put yourself in the shoes of a dictator…

…are you going to allow this? Nope!

🏡

🌎

Put yourself in the shoes of a dictator…

…can you really accomplish this? Yes, if you’re strategic about it…

🏡

🌎

What can dictators do about it?

transnational repression

/trænsnæʃɪnʌl rɛprɛsʃən/    noun


Attempts by (mostly) authoritarian states to control and silence their dissident populations abroad

What can we do to address TNR?

Existing data initiatives (Dukalskis et al. 2022)

  • Authoritarian Actions Abroad Database1
  • Freedom House’s global survey of TNR2
  • Central Asian Political Exiles Data Project
  • China’s Transnational Repression of Uyghurs

Obstacles to maintaining a comprehensive, up-to-date evidence base

  • Observational challenges: designed to be discreet, spans multiple territories/jurisdictions
  • Practical concerns: time/labor costs, shifts in government funding

AI-based methods for addressing these challenges

  • Domain models (e.g., ConfliBERT) for semantic retrieval; codebook-guided LLM extraction
  • Actor embeddings for complex attribution (Croicu and Maase 2025)

Project Aims

Research Questions

  • RQ1: How is TNR represented in public reporting?
  • RQ2: What is the global scope of TNR?
  • RQ3: How has TNR evolved over the last 35 years?

Objectives

  • Build a measurement pipeline tailored to the multi-juridsictional and discreet nature of TNR
  • Expand existing datasets using our pipeline to estimate global scope
  • Map temporal shifts, especially as they relate to digital technologies and global migration

Links to EWS

  • “Canary in the coal mine” for regime change and risks to politically threatening populations

RQ1: How is TNR represented in public reporting?

Actors Involved

Perpetrating/origin states, target/host states, media outlets, interstate relationships

  • Multiple actors and states ➡️ more variation in how events are reported

General Expectations

  • Low recall: TNR is designed to be hidden ➡️ we will probably find a lot of false negatives
  • We expect to find higher levels of attributional uncertainty in public reporting on TNR
    • Variations in attributional uncertainty based on political/geographic proximity to the perpetrating state and whether the reporting outlet is within the perpetrator-host dyad

RQ1: Empirical Approach

Data

  • Search procedure from the AAAD Codebook ➡️ corpus of potentially relevant article texts
  • Article texts will be scraped via Google Custom Search API, News API, and Lexis+ (1991–2025)
  • Trim texts to account for input limits without losing information variables of interest

Measurement

  • Identifying TNR articles
    • Run baseline tests with proprietary LLMs and transformer models
    • Identify whether we observe low recall and attributional uncertainty
    • Fine tuning, hypothesis testing
  • Measuring attributional uncertainty: Composite score based on text features that indicate passive language constructions and suppressing information on perpetrators

Takeaways

Why TNR?

  • ⬆️ geopolitical salience, especially with:
    • global democratic backsliding
    • shifts in migration
    • developments in digital technologies

Our approach

  • We plan to adapt existing methods from conflict prediction to track TNR:
    • Information extraction methods (transformer and/or foundation models)
    • More complex actor embeddings

Measurement obstacles

  • TNR is difficult to measure:
    • multiple territories and jurisdictions
    • perpetrators intentionally obscure attribution

Implications for EWS

  • Canary in the coalmine for:
    • predicting regime change/democratic backsliding
    • assessing risks for marginalized and/or politically threatening populations

References

Alonso, Andoni, and Pedro Oiarzabal. 2010. Diasporas in the New Media Age: Identity, Politics, and Community. University of Nevada Press.
Bernal, Victoria. 2020. “African Digital Diasporas: Technologies, Tactics, and Trends: Introduction.” African Diaspora 12 (1-2): 1–10.
Brinkerhoff, Jennifer M. 2009. Digital Diasporas: Identity and Transnational Engagement. Cambridge University Press.
Croicu, Mihai, and Simon Polichinel von der Maase. 2025. “From Newswire to Nexus: Using Text-Based Actor Embeddings and Transformer Networks to Forecast Conflict Dynamics.” arXiv Preprint arXiv:2501.03928.
Dukalskis, Alexander. 2021. Making the World Safe for Dictatorship. Oxford University Press.
Dukalskis, Alexander, Saipira Furstenberg, Yana Gorokhovskaia, John Heathershaw, Edward Lemon, and Nate Schenkkan. 2022. “Transnational Repression: Data Advances, Comparisons, and Challenges.” Political Research Exchange 4 (1): 2104651. https://doi.org/10.1080/2474736X.2022.2104651.
Esberg, Jane, and Alexandra Siegel. 2020. “How Exile Shapes Online Opposition: Evidence from Venezuela.”
Gorokhovskaia, Yana, and Isabel Linzer. 2022. “Defending Democracy in Exile: Policy Responses to Transnational Repression.” Report. Freedom House. https://freedomhouse.org/sites/default/files/2022-05/Complete_TransnationalRepressionReport2022_NEW_0.pdf.
Keck, Margaret E., and Kathryn Sikkink. 1999. “Transnational Advocacy Networks in International and Regional Politics.” International Social Science Journal 51 (159): 89–101. https://doi.org/10.1111/1468-2451.00179.
Michaelsen, Marcus. 2018. “Exit and Voice in a Digital Age: Iran’s Exiled Activists and the Authoritarian State.” Globalizations 15 (2): 248–64.
Vaughan, Grady, Yana Gorokhovskaia, and Nate Schenkkan. 2025. “Ten Findings from Ten Years of Data on Transnational Repression.” Freedom House Perspectives. https://freedomhouse.org/article/ten-findings-ten-years-data-transnational-repression.